Gen3R：三维场景生成与前馈重建的结合
Gen3R: 3D Scene Generation Meets Feed-Forward Reconstruction
把 VGGT 这类前馈重建模型当作几何编码器，用适配器把其 token 对齐到视频扩散模型的外观潜空间，从而在同一个扩散过程里同时生成 RGB 视频与相机位姿、深度、点云；对做 3D 生成的人，这是一条「复用重建基础模型当几何先验」的现成路径，对做重建的人则提示可用生成先验提升鲁棒性。
原文依据 We repurpose the VGGT reconstruction model to produce geometric latents by training an adapter on its tokens
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论本文提出 Gen3R，将基础重建模型的强先验与视频扩散模型相结合，用于场景级三维生成。我们改造 VGGT 重建模型：在其 token 上训练一个适配器以产生几何潜变量，并通过正则化使这些潜变量与预训练视频扩散模型的外观潜变量对齐。通过联合生成这些相互解耦却又彼此对齐的潜变量，Gen3R 同时输出 RGB 视频与对应的三维几何，包括相机位姿、深度图和全局点云。实验表明，本方法在单图与多图条件的三维场景生成上取得了最优结果。此外，本方法能够借助生成先验提升重建的鲁棒性，体现了紧耦合重建模型与生成模型所带来的相互增益。
英文原文摘要 876 字符
We present Gen3R, a method that bridges the strong priors of foundational reconstruction models and video diffusion models for scene-level 3D generation. We repurpose the VGGT reconstruction model to produce geometric latents by training an adapter on its tokens, which are regularized to align with the appearance latents of pre-trained video diffusion models. By jointly generating these disentangled yet aligned latents, Gen3R produces both RGB videos and corresponding 3D geometry, including camera poses, depth maps, and global point clouds. Experiments demonstrate that our approach achieves state-of-the-art results in single- and multi-image conditioned 3D scene generation. Additionally, our method can enhance the robustness of reconstruction by leveraging generative priors, demonstrating the mutual benefit of tightly coupling reconstruction and generative models.